Re: postgresql 8.0 with cygwin - success

Marek Lewczuk schrieb:
> Reini Urban wrote:
>> Bruce Momjian schrieb:
>>> Great. Thanks. I was getting conflicting reports of Cygwin problems
>>> before.
>>
>> Not for me:
>> I found a problem with my pg_ctl patch (WIN32 is always defined in
>> cygwin when windows.h is included, and the START /B postmaster cmdline
>> doesn't work on cygwin). Patch will be posted soon.
>
> Right. pg_ctl is not working - I didn't noticed that, because I use
> postmaster instead pg_ctl. Below pg_ctl error:
> $ pg_ctl start
> START: not found
> pg_ctl: could not start postmaster: exit code was 32512

Bruce:
> was wondering about that use of START. What I think we will do
>is to add a comment that including windows.h in that file defines WIN32
>and have an #ifndef __CYGWIN__ in places that need it --- just let me
>know.

Attached is my improved version: (just as you said)

* Fix previous pg_ctl patch (WIN32 is always defined in cygwin when
windows.h is included, and the START /B postmaster cmdline doesn't
work on cygwin.

You can safely ignore the patches of other two files, but these are my
remaining issues. (cygpg.dll just for overall conformancy).

I still have to test how it interacts with the eventlog if started as
service. (under different users)
However it should be better than the 7.x versions, which had lotta
problems with logfile file-permissions, because services are run as the
SYSTEM user.

>> But I finished now the plperl.sql regression tests and wait for
>> a clean regression suite to run through.
>
> It would be great - I've just tried to run some plperl function, but no
> luck (postgresql crashed).

:( as with 7.x

But first I have to find the SUBST problem (errno 140).
Then I will do some heavy IPC debugging with plperl.
(src/backend/port/ipc_test.c has to be fixed for old-style elog handling)
Then I have to do the mbregress tests (multibyte on cygwin? maybe newlib
has some remaining issues)
--
Reini Urban